Re: Crash while building xemacs 21.5-b9 (+CVS-20021017)

Thanks for the bug report, I've never seen anything quite like that. :-) Stephen> (1) configure with --pdump (2) configure with Stephen> --use-union-type (3) configure with --with-system-malloc Stephen> in various combinations. (1) and (2) on general Stephen> principles, (3) has been reported to help with Red Hat Stephen> 8.0/GCC 3.2 crashes. njsf> Thanks, (3) did the trick, it must be a gcc 3.2 with glib njsf> 2.3.1 issue. gcc 3.2 with glibc 2.2.5 was fine. njsf> Do you recommend using (1) and (2) ? (1) Yes, because GCC and binutils work recently has changed the layout of the executable, which the unexec dumper has unholy knowledge of. pdump does it by hand, so to speak, and therefore is not dependent on upstream changes (especially to ld). (2) A matter of taste. Backtraces are slightly more informative (copious amounts of new data, but not much extra information). It may be slightly slower. I don't actually _know_ of any platforms that need --use-union-type right now, while there are several platforms that simply don't have an unexec-style dumper, and several recent crashes that go away with --pdump. OTOH, the casting of int to Lisp_Object * and back that is characteristic of --use-union-type=no just looks risky for the future, as optimization gets more aggressive. -- Institute of Policy and Planning Sciences http://turnbull.sk.tsukuba.ac.jp University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN My nostalgia for Icon makes me forget about any of the bad things. I don't have much nostalgia for Perl, so its faults I remember. Scott Gilbert c.l.py